Sign In

Nightmare Grunge- Urban City Darkness, and Spooky Horror Nature Scene Enhancer

1.4k

Updated: May 29, 2026

stylehorrorcitynaturehalloweenspooky

Download

1 variant available

SafeTensor

324.99 MB

Verified:

Type

LoRA

Stats

189

1.4K

699

Reviews

Published

May 27, 2026

Base Model

Illustrious

Hash

AutoV2
BB9AA06908

Trigger Words

indoors
outdoors
city
grunge

Illustrious is great, except it is very "clean" and "bright"- grunge, dirt, darkness and atmospheric horror details are not its strong suit. This LoRA is designed to turn your bright, plastic-smooth and pretty images into gritty, grungy, horror-esque visions of urban decay and haunted nature. This style LoRA enhances edgy grunge characters, zombie apocalypse cities, eerie Halloween wetlands, Eldritch cursed shores, grimy apartment interiors, and more!

Please read below for important usage details:

Trigger Word: Please start you prompt with either the indoors, outdoors, city or grunge tag, this will activate the LoRA and help guide the image depending on which kind of scene you want. Add the tags night, grunge, dirty, damaged and/or gritty texture as desired to ensure consistent results.

LoRA Strength- use 1 (full). You can go lower, but do not go over. Full strength keeps the style looking its best, but lower it a little if a certain concept isn't working. I had to use 0.8 strength to get raining weather to work (I'll remember that for a future update).

Sampler: DPM++ 2M SDE is strongly recommended, or DPM++ 2M (cheaper onsite), but avoid Euler A! I normally prefer Euler A, but on this it "smooths" details too much and tends to look bad.

All showcase images were made using the *Plant Milk Model Suite- Walnut" checkpoint. I linked it here under "suggested resources". I really recommend it, it's my favorite, and works best with this, but feel free to experiment.

Because Hires-Fix will try to "clean" the roughness out of this I recommend skipping the Hires Fix step entirely. Even just 0.1 denoising strength removes some noticeable grit details. You can improve image quality by generating at a higher (SDXL supported) base resolution, adding more sampling steps, or using adetailer/face-fix options, but a full image-to-image Highres fix will scrub some of the "dirty" details away. Please look closely at the comparison below for an example:

Finally, I want to thank my friends, too many to name, for all of their help testing and getting this model ready- I love you all! Feel free to peruse the showcase images and gallery for ideas and example, and if you find the model helpful, please consider leaving a review with an image post- it really helps me out 💕